“Editing after machine translation”… a phrase that has become commonplace in the world of languages!
Undoubtedly, technological advancements have impacted various sectors worldwide, and artificial intelligence, with its technologies, has permeated all fields, bringing about numerous radical changes. Development is no longer a mere byproduct of our lives; it has become the primary driver of trends. The reality is no different in the world of translation and languages.
Before delving deeper into the implications of artificial intelligence on translation mechanisms and how languages are used in everyday and professional settings, let’s first understand what artificial intelligence actually is.
How artificial intelligence works in the field of translation
In recent years, artificial intelligence has developed significantly in the field of translation, and reliance on these technologies is increasing noticeably.
In fact, machine translation is not a new thing; we have known it for many years, especially through well-known sites such as Google Translate and Microsoft Translator.
In general, AI-based translation technologies undergo intensive training to gain a deeper understanding of the language, as their concept is based on learning from repetition, exploring errors, and then correcting them to arrive each time at a better version of the final text.
Some of the features of AI-based translation can be discussed, most notably:
The ability to analyze massive amounts of data, i.e., translate large texts in record time;
the ability to translate into all languages, including rare languages with limited specialists;
lower translation costs compared to human translation
; and the ability to process various types of media and files, whether audio, video, or others.
But is the picture really as “rosy” as it seems?
Challenges of AI-based translation… Why is abandoning human capabilities impossible?
The advanced level of development we’ve reached today might paint a rosy, almost perfect picture of the translation world, especially for clients who are always looking for quick and low-cost solutions. But reality differs from appearances. So, what challenges should we highlight?
Literal translation: Literal translation is one of the most discussed problems when relying on artificial intelligence to translate from one language to another. The technologies used are unable to consider or understand the general context of the text to be translated, and this leads to a loss of the text’s specificity – which could be a scientific article, a consulting presentation, a legal document, or something else – and to vague and inaccurate phrases, and even to the use of “incorrect” terms, that is, terms different from the correct specialized terms.
Uncreative Translation: No matter how much computer scientists and technical specialists train translation systems and software, they will never be able to replicate the creativity that the human mind, and more importantly, the human heart, can produce. Machine translation is, without a doubt, a rigid and emotionless process. We’re not necessarily talking about literary texts here; we can take an example from the job market, specifically the files that consulting firms prepare for their clients.
Recently, we’ve seen numerous posts on LinkedIn discussing consulting translation. These posts highlight the difficulty of satisfying clients in some cases, even when using the correct vocabulary and providing a translation that meets linguistic and semantic standards. The challenge lies in conveying the idea with utmost precision—given that the consultant possesses deeper and more specialized knowledge of the subject matter—and in delivering the presentation with the same persuasive style, or what we call “tone of voice.”
Based on this, it can be emphasized once again the importance of human translation, which is superior to machine translation in this context, given the translator’s ability to understand the interaction that the client is aiming for in both the source and target languages.
Difficulty in understanding dialects: This problem is particularly evident during the processing of video and audio clips, as the artificial intelligence system may misunderstand and mispronounce some terms, leading to errors that can be avoided by using a human translator.
Privacy: Sharing information on AI-based translation systems and software means a loss of data privacy, as the information becomes indirectly accessible for other purposes. This problem can only be avoided by using a dedicated, company-owned server, a solution that incurs significant costs.
Post-machine translation editing… what do we mean by it?
Post-machine translation editing is emerging as one of the latest trends in the field of translation. Post-machine translation editing, or editing after machine translation, is the process of improving machine-translated text using human expertise to review the content and make necessary adjustments.
This trend has gained considerable popularity and widespread adoption as a viable solution for a large number of clients seeking quick and low-cost solutions. For them, it allows them to bypass the time-consuming and laborious process of translation and simply hire an editor to review the machine-translated text, correct it, and make a limited number of adjustments.
But what challenges might a translator face when editing machine translation that the client is unaware of?
Post-machine translation editing… a process not without its challenges
Some might think that artificial intelligence handles most of the work, leaving the human translator with only the easier part. However, this isn’t true!
Translators face numerous challenges and may sometimes need to double their efforts to produce a final translated text that satisfies the client and meets their expectations. So, what challenges are we referring to?
Inconsistency in key terms: Artificial intelligence is unable to standardize key terms during translation, leading to overall inconsistency across different parts of the document. In such cases, the human translator focuses all their efforts on achieving a consistent translation that ensures smooth comprehension.
Failure to adhere to references : Sometimes, the client provides the translator with reference texts that they can use to ensure the correct terminology aligns with the preferences of the client or organization. Similarly, the translator often consults official websites to select vocabulary and sentence structure that conform to the client’s preferred style. Artificial intelligence, which provides translation suggestions based on its training and the data available in its central database, lacks these capabilities.
Obtaining out-of-context translations : As mentioned earlier, AI technologies sometimes fail to consider the overall context of the text, resulting in inappropriate translations. For example, when searching for “Stakeholders Ecosystem,” we might get “النظام البيئة أصحاب التخصصي” (the ecosystem of stakeholders). However, the correct translation might be “نظام أصحاب التخصصي” (stakeholder system) or “بيئة أصحاب التخصصي” (stakeholder environment), depending on the context. Therefore, a human translator must carefully review the phrases and terms suggested by the automated system to ensure they fit the context.
Abbreviations : Abbreviations present an additional challenge in the post-machine translation editing process. Three main problems can be identified: first, leaving abbreviations untranslated – even in simple cases (such as the abbreviation for a country’s name: KSA); second, translating abbreviations incorrectly without any connection to the text’s context or subject matter; and third, presenting the same abbreviation with different translations. All these problems require extra attention from the human translator to ensure accurate and consistent translation.
Additions and Omissions : It is no secret to those working in and specializing in translation that artificial intelligence sometimes tends to add ideas and phrases not present in the original or source text, or even to omit key concepts that should be emphasized. In certain specific contexts, this is a fundamental problem that must be addressed.
Summary
Can human capabilities be dispensed with and replaced in translation? No. Can
artificial intelligence be utilized to some extent? Yes .
Based on all of the above, it can be affirmed that no matter how advanced artificial intelligence technologies become, they will never be able to surpass humans, whether in the field of translation or any other field. The human mind possesses capabilities that ensure a deep understanding of the text, consideration of the overall context, and the provision of correct and accurate linguistic options. However, this does not preclude the use of artificial intelligence and machine translation programs to explore more options and suggestions, and thus arrive at a polished translated text.
No matter how much the quality of machine translation improves and how reliable its results become, it will not be able to overcome the challenges we have discussed, most notably preserving the context, adhering to the “tone” of the original author’s voice, and considering the client’s identity and preferred style.